Research map / Probabilistic and scientific modeling / Causal discovery
Granger causal inference on DAGs identifies genomic loci regulating transcription
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DAG-based lagged message passing rather than correlation-based ranking | GrID-Net · sci-CAR, SHARE-seq, and SNARE-seq eQTL prediction | AUPRC: 0.0594 to 0.0645 $\pm$ 0.0004 (sci-CAR); 0.0385 to 0.0604 $\pm$ 0.0041 (SHARE-seq); 0.0096 to 0.0198 $\pm$ 0.0088 (SNARE | Helped |
| use DAG-based lagged message passing rather than correlation-based ranking | GrID-Net · sci-CAR, SHARE-seq, and SNARE-seq chromatin interaction prediction | AUROC: 0.509 $\pm$ 0.005 to 0.550 $\pm$ 0.001 (sci-CAR); 0.506 to 0.753 $\pm$ 0.010 (SHARE-seq); 0.517 to 0.540 $\pm$ 0.006 (SN | Helped |
| replace diffusion pseudotime with Palantir or Monocle (ablation) | GrID-Net · eQTL prediction across three datasets | AUPRC: Diffusion: $ 0.065 \pm 0.0004$, $ 0.060 \pm 0.0041$, $ 0.020 \pm 0.0088$; Palantir: $ 0.064 \pm 0.0004$, $ 0.060 \pm 0.0 | No clear effect |
| increase kNN graph size from 15 to 30 or 50 (ablation) | GrID-Net · eQTL prediction across three datasets | AUPRC: sci-CAR: 0.0645 $\pm$ 0.0004 for k=30, k=50, and k=15 | No clear effect |
| use Welch's t-test instead of F-test to rank candidate interactions (ablation) | GrID-Net · eQTL prediction | AUPRC: 0.0645 $\pm$ 0.0004 to 0.0697 $\pm$ 0.0006 (sci-CAR); 0.0604 $\pm$ 0.0041 to 0.0573 $\pm$ 0.0047 (SHARE-seq); 0.0198 $\p | Mixed |
| train full and reduced models separately instead of with a combined loss (ablation) | GrID-Net · eQTL and chromatin interaction prediction | AUPRC: eQTL combined/separated: sci-CAR 0.0645 $\pm$ 0.0004 / 0.0645 $\pm$ 0.0004; SHARE-seq 0.0604 $\pm$ 0.0041 / 0.0603 $\pm$ | No clear effect |
| increase GNN depth to allow greater pseudotime lag (ablation) | GrID-Net · top 1% peak--gene pairs across three datasets | distal-to-proximal ratio: sci-CAR: 6.10 $\pm$ 0.062 (2 layers), 6.23 $\pm$ 0.119 (5), 6.05 $\pm$ 0.055 (10); SHARE-seq: 12.53 $\pm$ 0.128, 15.33 $ | Mixed |
| aggregate sparse observations through the DAG rather than use existing methods | GrID-Net · sparse-gene and sparse-peak eQTL prediction | AUPRC: SHARE-seq sparse peaks: 0.0261 $\pm$ 0.0040 (GVAR) to 0.1450 $\pm$ 0.0310 (GrID-Net) | Helped |
Closest papers by meaning
- Gene Regulatory Network Inference in the Presence of Selection Bias and Latent Confounders (NeurIPS 2025 · similarity 0.87)
- Gene Regulatory Network Inference in the Presence of Dropouts: a Causal View (ICLR 2024 · similarity 0.87)
- Scalable and Flexible Causal Discovery with an Efficient Test for Adjacency (ICML 2024 · similarity 0.87)
- KINDLE: Knowledge-Guided Distillation for Prior-Free Gene Regulatory Network Inference (NeurIPS 2025 · similarity 0.86)
- Interventional Causal Discovery in a Mixture of DAGs (NeurIPS 2024 · similarity 0.86)
- Identifying biological perturbation targets through causal differential networks (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.